Test for glyphs which straddle the edge of device space.

Test for glyphs which straddle the edge of device space.

If a glyph ends up with a left edge near the extreme right edge of
device space, the right edge may end up at the extreme left. The
current clip tests do not test for this situation and attempt to
draw the glyphs. This results in the blitters never reaching the
right edge of the glyph and so reading from other memory.

BUG=chromium:467011
Committed: https://skia.googlesource.com/skia/+/d095f2bf06ca810844233bea9fb06de585643b32

[bungeman-skia, 2015-03-16 15:17:02]

bungeman@google.com changed reviewers:
+ reed@google.com

[bungeman-skia, 2015-03-16 15:17:02]

While we're trying to figure out what might be the 'real' fix, this is a trivial
fix which gets things working for now and the fuzzer from complaining (it's
opened a dozen bugs over the weekend around this).

[bungeman-skia, 2015-03-16 16:25:25]

PS2 does a full overflow/clamp check up front.

[reed1, 2015-03-16 16:39:28]

possible additional refactoring...

1. (unsigned)fx < conservative_max
2. capture the test as a stand-alone predicate function
3. move to caller, so all of our procs will get the check (and their signatures
could revert back to 16.16)

[reed1, 2015-03-16 16:39:50]

if you want to land early, lgtm but lets add a unittest to exercise this.

[bungeman-skia, 2015-03-16 16:42:59]

The CQ bit was checked by bungeman@google.com

[commit-bot: I haz the power, 2015-03-16 16:43:44]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1012763002/20001

[commit-bot: I haz the power, 2015-03-16 16:49:59]

Committed patchset #2 (id:20001) as
https://skia.googlesource.com/skia/+/d095f2bf06ca810844233bea9fb06de585643b32